good point, formal support for merging paper/model space should be a wish list item, thanks for that.

Your terminology is pretty much dead on. There are a few choices with respect to xrefs today in PowerCADD:

1. Translations of each Xref drawing could be copied and pasted into a symbol library and inserted later into any powercadd drawing. The primary benefit of PowerCADD Symbol Libraries being they are drawing independent. That works but libraries currently do not support layers so it may not be practical depending on the drawing content. To be of tangible benefit in a symbol library (as they exist today but who knows about the future), using DWGi files that already represent typical blocks for simplier components is easiest (note I don't say 'best') and provides the most flexibility. I really should write a tutorial now to best use Symbols now that you mention this subject
2. Referencing: Yes, PowerCADD can reference any type of file format that is currently supports. That includes a DWG file. The primary issue with that approach is if a DWG file is inserted as a reference, the Translation Dialog appears every time the referenced dwg file is loaded in pcadd which is generally inconvenient. However, it could be a useful approach if there were frequent enough revisions in ACad to warrant dealing with the dialog every time. Otherwise it's generally more efficient to translate the DWG file, save it as a PCadd file, and reference that. If it changes at a later date just do the translation again and replace the PCADD file that's referenced
3. If each DWG file is translated, then referenced as a PCADD file, that has pro's and con's. The pro being each referenced file can be updated separately. If a new dwg comes in that replaces just one reference file, it's a single translation and replacement in pcadd. If each referenced file contains a common element (say a site grid or architectural grid) then getting to line up and being confident they're lined up is a virtual no brainer and the process would work great.
4. The down side is getting things to line up as noted below.

The fundamental problem with any of the above methods, be it referencing or inserting drawings as symbols, is the issue of coordination location and getting the xref translations to line up correctly in PowerCADD. While they are lined up correctly in AutoCAD coordinate space, they are not necessarily matched up once translated. I say it that way since I have successfully translated multiple xref's and had them line up correctly -- but that's the exception rather than the rule. Typically it's related to how the original AutoCAD files were setup and/or how the translation settings are specified in PowerCADD. It takes a very skilled hand at both ends to get that to work which is typically not a safe wager.

It's just not safe, in my mind, to trust that a separate translation of each Xref drawing can be successfully aligned after the fact. The risk of an error just seems to great to warrant the attempt so having the drawing bound in ACad first makes that process 'safer' from a cover my arse point of view. Binding the file isn't hard in Acad as you know but it is an extra step that some consultants are reluctant to do. In those cases, I typically use AutoCAD to bind them myself, then do the PowerCADD translation after that. Having just said that -- one can see the issues that Engineered Software have to address and their risk and time required to get the feature request we're discussing here to work correctly and be confident in their translation process.

Once translated, a DWG file (be it stand alone or xref'd) can be used as a distinct PowerCADD drawing, as a symbol in a library, or referenced into other PCadd drawings. The type of content and the ultimate end use and frequency of change really dictates how they should be managed and which one of those options should be used.

I hope that makes some sense
Brian